## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

You've reached the end of this preview. Sign up to read more!

Page 1 of 1

**Handbook of Global Analysis **

Demeter Krupka

David Saunders

*Palacky University, Olomouc, Czech Republic *

**Cover image **

**Title page **

**Copyright **

**Preface **

**Chapter 1: Global aspects of Finsler geometry **

**1 Finsler metrics and connections **

**2 Geodesics in Finsler manifolds **

**3 Comparison theorems: Cartan—Hadamard theorem, Bonnet—Myers theorem, Laplacian and volume comparison **

**4 Rigidity theorems: Finsler manifolds of scalar curvature and locally symmetric Finsler metrics **

**5 Closed geodesics on Finsler manifolds, sphere theorem and the Gauss—Bonnet formula **

**Chapter 2: Morse theory and nonlinear differential equations **

**1 Introduction **

**2 Global theory **

**3 Local theory **

**4 Applications **

**5 Strongly indefinite Morse theory **

**6 Strongly indefinite variational problems **

**Chapter 3: Index theory **

**1 Introduction and some history **

**2 Fredholm operators – theory and examples **

**3 The space of Fredholm operators and K-theory **

**4 Elliptic operators and Sobolev spaces **

**5 The Atiyah-Singer Index Theorem **

**6 Generalizations **

**Chapter 4: Partial differential equations on closed and open manifolds **

**1 Introduction **

**2 Sobolev spaces **

**3 Non-linear Sobolev structures **

**4 Self-adjoint linear differential operators on manifolds and their spectral theory **

**5 The spectral value zero **

**6 The heat equation, the heat kernel and the heat flow **

**7 The wave equation, its Hamiltonian approach and completeness **

**8 Index theory on open manifolds **

**9 The continuity method for non-linear PDEs on open manifolds **

**10 Teichmüller theory **

**11 Harmonic maps **

**12 Non-linear field theories **

**13 Gauge theory **

**14 Fluid dynamics **

**15 The Ricci flow **

**Chapter 5: The spectral geometry of operators of Dirac and Laplace type **

**1 Introduction **

**2 The geometry of operators of Laplace and Dirac type **

**3 Heat trace asymptotics for closed manifolds **

**4 Hearing the shape of a drum **

**5 Heat trace asymptotics of manifolds with boundary **

**6 Heat trace asymptotics and index theory **

**7 Heat content asymptotics **

**8 Heat content with source terms **

**9 Time dependent phenomena **

**10 Spectral boundary conditions **

**11 Operators which are not of Laplace type **

**12 The spectral geometry of Riemannian submersions **

**Acknowledgments **

**Chapter 6: Lagrangian formalism on Grassmann manifolds **

**1 Introduction **

**2 Grassmann manifolds **

**3 The Lagrangian formalism on a Grassmann manifold **

**4 Lagrangian formalism on second order Grassmann bundles **

**5 Some applications **

**Acknowledgements **

**Chapter 7: Sobolev spaces on manifolds **

**1 Motivations **

**2 Sobolev spaces on manifolds: definition and first properties **

**3 Equality and density issues **

**4 Embedding theorems, Part I **

**5 Euclidean type inequalities **

**6 Embedding theorems, Part II **

**7 Embedding theorems, Part III **

**8 Compact embeddings **

**9 Best constants **

**10 Explicit sharp inequalities **

**11 The Cartan-Hadamard conjecture **

**Chapter 8: Harmonic maps Dedicated to the memory of James Eells **

**Introduction **

**1 Harmonic functions on Euclidean spaces **

**2 Harmonic maps between Riemannian manifolds **

**3 Weakly harmonic maps and Sobolev spaces between manifolds **

**4 Regularity **

**5 Existence methods **

**6 Other analytical properties **

**7 Twistor theory and completely integrable systems **

**Chapter 9: Topology of differentiable mappings **

**1 Introduction **

**2 Manifolds and singularities **

**3 Milnor fibre **

**4 Monodromy **

**5 Stratifications of spaces **

**6 Stratified Morse theory **

**7 Rectified homotopical depth **

**8 Relative stratified Morse theory **

**9 Topology of images and multiple point spaces **

**10 The image computing spectral sequence **

**Chapter 10: Group actions and Hilbert’s fifth problem **

**Introduction **

**1 Hilbert’s fifth problem **

**2 Lie groups and manifolds **

**3 Group actions **

**4 Cartan and proper actions of Lie groups **

**5 Non-paracompact Cartan G-manifolds **

**6 Homogeneous spaces of Lie groups **

**7 Twisted products **

**8 Slices **

**9 The strong Cr topologies, 1 ≤ r ≤ ∞, and the very-strong C∞ topology **

**10 Continuity of induced maps in the strong Cr topologies, 1 ≤ r < ∞, and in the very-strong C∞ topology **

**11 The product theorem **

**12 The equivariant glueing lemma **

**13 Whitney approximation **

**14 Haar integrals of Cs maps, 1 ≤ s ≤ ∞, and of real analytic maps **

**15 Continuity of the averaging maps in the strong Cr topologies, 1 ≤ r < ∞, and in the very-strong C∞ topology **

**16 Approximation of K-equivariant Cs maps, 1 ≤ s ≤ ∞, by K-equivariant real analytic maps, in the strong Cs topologies, 1 ≤ s < ∞, and in the very-strong C∞ topology **

**17 Approximation of Cs K-slices, 1 ≤ s ≤ ∞ **

**18 Proof of the main theorem **

**Chapter 11: Exterior differential systems **

**1 Introduction **

**2 Exterior differential systems **

**3 Basic existence theorems for integral manifolds of C∞ systems **

**4 Involutive analytic systems and the Cartan-Kähler Theorem **

**5 Prolongation and the Cartan-Kuranishi Theorem **

**6 A Cartan-Kähler Theorem for C∞ Pfaffian systems **

**7 Characteristic cohomology **

**8 Topological obstructions **

**9 Applications to second-order scalar hyperbolic partial differential equations in the plane **

**10 Some applications to differential geometry **

**Chapter 12: Weil bundles as generalized jet spaces **

**Preface **

**1 Weil algebras **

**2 Weil bundles **

**3 On the geometry of TA-prolongations **

**4 Fiber product preserving bundle functors **

**5 Some applications **

**Chapter 13: Distributions, vector distributions, and immersions of manifolds in Euclidean spaces **

**1 Introduction **

**2 Distributions on Euclidean spaces **

**3 Distributions and related concepts on manifolds **

**4 Vector distributions or plane fields on manifolds **

**5 Immersions and embeddings of manifolds in Euclidean spaces **

**Chapter 14: Geometry of differential equations **

**Introduction **

**1 Geometry of jet spaces **

**2 Algebra of differential operators **

**3 Formal theory of PDEs **

**4 Local and global aspects **

**Chapter 15: Global variational theory in fibred spaces **

**1 Introduction **

**2 Prolongations of fibred manifolds **

**3 Differential forms on prolongations of fibred manifolds **

**4 Lagrange structures **

**5 The structure of the Euler-Lagrange mapping **

**6 Invariant variational principles **

**7 Remarks **

**Acknowledgement **

**Chapter 16: Second Order Ordinary Differential Equations in Jet Bundles and the Inverse Problem of the Calculus of Variations **

**1 Introduction **

**2 Second-order differential equations on fibred manifolds **

**3 Variational structures in the theory of sodes **

**4 Symmetries and first integrals **

**5 Geometry of regular sodes on × TM **

**6 The inverse problem for semisprays **

**Acknowledgements. **

**Chapter 17: Elements of noncommutative geometry **

**1 Introduction **

**2 Algebras instead of spaces **

**3 Modules as bundles **

**4 Homology and cohomology **

**5 The Chern characters **

**6 Connections and gauge transformations **

**7 Noncommutative manifolds **

**8 Toric noncommutative manifolds **

**9 The spectral geometry of the quantum group SUq(2) **

**Chapter 18: De Rham cohomology **

**Introduction **

**1 De Rham complex **

**2 Integration and de Rham cohomology. De Rham currents. Harmonic forms [20] **

**3 Generalizations of the de Rham complex **

**4 Equivariant de Rham cohomology **

**5 Complexes of differential forms associated to differential geometric structures **

**Chapter 19: Topology of manifolds with corners **

**1 Introduction **

**2 Quadrants **

**3 Differentiation theories **

**4 Manifolds with corners **

**5 Manifolds with generalized boundary **

**Chapter 20: Jet manifolds and natural bundles **

**1 Introduction **

**2 Jets **

**3 Differential equations **

**4 The calculus of variations **

**5 Natural bundles **

**Chapter 21: Some aspects of differential theories Respectfully dedicated to the memory of Serge Lang **

**Introduction **

**1 Background **

**2 Calculus in topological vector spaces and beyond **

**3 The Chern – Rund derivative **

**Chapter 22: Variational sequences **

**Introduction **

**1 Preliminaries **

**2 Contact forms **

**3 Variational bicomplex and variational sequence **

**4 C-spectral sequence and variational sequence **

**5 Finite order variational sequence **

**6 Special topics **

**7 Notes on the development of the subject **

**Appendix: splitting the exterior algebra **

**Chapter 23: The Oka-Grauert-Gromov principle for holomorphic bundles **

**Introduction **

**1 Stein manifolds and Stein spaces **

**2 Oka’s theorem **

**3 Grauert’s Oka principle **

**4 Gromov’s Oka principle **

**5 The case of Riemann surfaces **

**6 Complete intersections **

**7 Embedding dimensions of Stein spaces **

**8 Oka principle with growth condition **

**9 Oka’s principle and the Moving Lemma in hyperbolic geometry **

**10 The algebraic version of Oka’s principle **

**Abstracts **

**Index **

Elsevier

Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands

Linacre House, Jordan Hill, Oxford OX2 8DP, UK

First edition 2008

Copyright © 2008 Elsevier B.V. All rights reserved

No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher

Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: **permissions@elsevier.com. Alternatively you can submit your request online by visiting the Elsevier web site at http://www.elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material **

**Notice **

No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made

**Library of Congress Cataloging-in-Publication Data **

A catalog record for this book is available from the Library of Congress

**British Library Cataloguing in Publication Data **

A catalogue record for this book is available from the British Library

ISBN: 978-0-444-52833-9

For information on all Elsevier publications visit our website at **books.elsevier.com **

Printed and bound in The Netherlands

08 09 10 11 12 10 9 8 7 6 5 4 3 2 1

The group of topics known broadly as ‘Global Analysis’ has developed considerably over the past twenty years, to such an extent that workers in one area may sometimes be unaware of relevant results from an adjacent area. The many variations in notation and terminology add to the difficulty of comparing one branch of the subject with another.

Our purpose in preparing this Handbook has been to try to overcome these difficulties by presenting a collection of articles which, together, give an overall survey of the subject. We have been guided in this task by the MSC2000 classification, and so the scope of the Handbook may be described by saying that it covers the 58-XX part of the classification: ranging from the structure of manifolds, through the vast area of partial differential equations, to particular topics with their own distinctive flavour such as holomorphic bundles, harmonic maps, variational calculus and non-commutative geometry. The coverage is not complete, but we hope that it is sufficiently broad to provide a useful reference for researchers throughout global analysis, and that it will also be of benefit to mathematical physicists and to PhD and post-doctoral students in both areas.

The main work involved in the preparation of the Handbook has, of course, been that of the authors of the articles, who have carried out their task with skill and professionalism. Our debt to them is immediate and obvious. Some other potential authors have, for personal reasons, been unable to offer contributions to the Handbook, but we hope that those omissions will not detract too much from its value. The editors also wish to acknowledge the assistance of Petr Volný in the formatting of the LATEX manuscripts, and of Andy Deelen, Kristi Green and Simon Pepping at Elsevier for their help and advice during the preparation of the book. In addition, we should like to record our particular thanks to Arjen Sevenster from Elsevier, who commissioned the project and gave us support and encouragement during its development.

We should finally like to acknowledge the support of the Czech Science Foundation (grant 201/06/0922) and the Czech Ministry of Education, Youth, and Sports (grant MSM 6198959214) for our work on this project.

**Demeter Krupka and David Saunders, ***Palacký University, Olomouc. *

**Tadashi Aikou, Department of Mathematics and Computer Science, Kagoshima University, 1-2-35 Korimoto, Kagoshima, 890 Japan. E-mail address: aikou@sci.kagoshima-u.ac.jp **

**Laszlo Kozma, Institute of Mathematics, University of Debrecen, H-4010 Debrecen, P O Box 12, Hungary. E-mail address: kozma@math.unideb.hu **

1 Finsler metrics and connections

2 Geodesics in Finsler manifolds

3 Comparison theorems: Cartan-Hadamard theorem, Bonnet-Myers theorem, Laplacian and volume comparison

4 Rigidity theorems: Finsler manifolds of scalar curvature and locally symmetric Finsler metrics

5 Closed geodesics on Finsler manifolds, sphere theorem and the Gauss-Bonnet formula

Let π : *TM *→ *M *be the tangent bundle of a connected smooth manifold *M *of dim *M *= *n*. We denote by *v *= (*x,y*) the points in *TM *if *y *∈ π−1(*x*) = *TxM*. We denote by *z*(*M*) the zero section of *TM*, and by *TM*x the slit tangent bundle *TM*\*z*(*M*). We introduce a coordinate system on *TM *as follows. Let *U *⊂ *M *be an open set with local coordinate (*x*¹, …, *xn*). By setting *v *= Σ*yi *(∂/∂*xi*)*x *for every *v *∈ π−1(*U*), we introduce a local coordinate (*x, y*) = (*x*¹, …, *xn*, *y*¹, …, *yn*) on π−1(*U*).

A function *F *: *TM *is called a *Finsler metric *on *M *if

(1) *F*(*x, y*) ≥ 0, and *F*(*x, y*) = 0 if and only if *y *= 0,

(2) *F*(*x*, λ*y*) = λ*F*(*x, y*: λ>0},

(3) *F*(*x, y*) is smooth on *TM*x, the out-side of the zero section,

(4) *G *= *F*²/2 is *strictly convex *on each tangent space *TxM*, that is, the Hessian (*Gij*) defined by

**(1.1) **

is positive-definite,

are satisfied. The pair (*M, F*) is called a *Finsler manifold*.

We note that the last condition in this definition is equivalent to the convexity of the unit ball *Bx *= {*y *∈ *TxM *| *F*(*x,y*) ≤ 1}.

*y *of each *y *∈ *TxM **y *= *F*(*x,y*), and the length *s*(*t*) of a smooth curve *c*(*t*) = (*x*¹(*t*),…, *xn*(*t*)) is defined by

.

(*Funk metric*) Let *g *be a Riemannian metric on *M*. We define α : *TM *. Since α is convex, there exists a 1-from β such that β(*v*) ≤ α(*v*). The function *F *= α + β defines a convex Finsler metric on *M *so-called *Randers metric*. We shall review a typical example of Randers metric (see *n *be an *n*-dimensional Euclidean space with the standard coordinate (*x*¹, …, *xn*), and B the unit ball centered the origin: B = {*x **n *| φ(*x**x *²>0}. The Riemannian metric *gH *defined by

is called the *Hilbert metric *on B. We define a 1-form β by

*H *of β with respect to *gH *β(*x**H **x *<1, and thus the function *F *is a Finsler metric called the *Funk metric *on B. We note that the relation between *gH *and *F *is given by

for all *v *∈ *TM*.

For the differential π* of the submersion π : *TM*x → *M*, the *vertical subbundle V *of *T*(*TM*x) is defined by *V *= ker π*, and *V *is locally spanned by {∂/∂*y*¹, …, ∂/∂*yn*} on each π−1 (*U*). Then it induces the exact sequence

where

is the pull-back bundle π**TM*.

Since the natural local frame field {∂/∂*xi*}*i*=1 …,*n *on each *U *on π−1(*U*), any section *X *is written in the form *X *= Σ(∂/∂*xi**Xi *for smooth functions *Xi *on each π−1 (*U*). Furthermore, since ker π* = *V*, the differential π* is given by π* = Σ(∂/∂*xi**dxi*.

We define a metric *G *by

**(1.3) **

for every section *X *= Σ(∂/∂*xi**Xi *and *Y *= Σ(∂/∂*xj**Yj*. We also set

by

**(1.4) **

for all sections *X, Y, Z *. It is trivial that *C *vanishes identically if and only if *G *is a Riemannian metric on *M*. This tensor field *C *is called the *Cartan tensor field*.

In the sequel, we use the notation *A*is naturally identified with *V *≅ ker π*, any section *X *is considered as a section of *V*. We denote by *XV *the section of *V *:

The following is trivial since **(1.2) is exact. **

**(1.5) **

.

+ ≅ {*cI *∈ *GL*(*n*); *c *+} ⊂ *GL*(*n*) acts on the total space by multiplication

+. This action induces a canonical section ε of *V *defined by ε(*v*) = (*v, v*) for all *v *∈ *TM*×. By the homogeneity of *F*, we have

, and we denote it by the same notation ∈, that is, ∈(*x, y*) = Σ(∂/∂*xi**yi*. This section ∈ is called the *tautological section of *and

**(1.6) **

For the submersion π : *TM*× → *M*, the vertical subbundle is defined by *V *= ker π*, while the *horizontal subbundle H *is defined by a subbundle *H *⊂ *T*(*TM*×) which is complementary to *V*. These subbundles give a smooth splitting

**(1.7) **

Although the vertical subbundle *V *is uniquely determined, the horizontal subbundle is not canonically determined. An *Ehresmann connection *of the submersion π : *TM*× → *M *is a selection of horizontal subbundles. In this report, we shall define this as follows.

An *Ehresmann connection *of the submersion π : *TM*× → *M *satisfying

**(1.8) **

.

If an Ehresmann connection θ is given, then a horizontal subbundle *H *is defined by *H *= ker θ. In this report, we shall assume that the subbundle *H *defined by θ is invariant by the action *m*•, that is, (*m*λ)**H *= *H *o *m*+. This assumption is equivalent to

**(1.9) **

**Remark 1.4: **A *linear connection *of the tangent bundle *TM *is a selection of horizontal subbundles in *GL*(*n*)-invariant way. Thus, an Ehresmann connection θ in our sense is sometimes called a *non-linear connection *of *TM*.

In the sequel, we denote by *Ak *the space of smooth *k*-valued *k*-form on *TM*× respectively. We suppose that an Ehresmann connection θ is given. Then, the exterior differential *d *: *Ak *→ *Ak*+1 is decomposed into the form *d *= *dH **dV *according to the decomposition **(1.7), where dH is the differential along H and dV is the one along Vis also decomposed into the form D = DH DV. **

*If an Ehresmann connection θ is given, then there exists a covariant exterior derivation D of **satisfying *

**(1.10) **

*or equivalently *

**(1.11) **

**Proof: **We define a covariant derivation *D *by *DV *= *dV *. It is easily shown that *D *= *DH **dV *. Then we have

and, from **(1.9) we obtain **

Therefore we obtain **(1.11). **

is also naturally identified with the horizontal subbundle *H*, and any section *X *is considered as a section of *H*. We denote by *XH *the section of *H *:

where

denotes the horizontal lift of natural local frame field {∂/∂*x*¹, …, ∂/∂*xn*} with respect to the given Ehresmann connection θ. The set {*dx*¹, …, *dxn*} is the dual basis of *H**. For the two bundle morphism π* and θ from *T*(*TM*, we have

*The bundle morphisms *π* *and θ satisfy *

**(1.12) **

*and *

**(1.13) **

*for every X *∈ *A*).

If a Finsler metric *F *is given on *TM*, then there exists a natural metric *G *defined by **(1.3). Then we shall introduce a covariant derivation ∇ which satisfies some natural axioms. **

by

**(1.14) **

With respect to the splitting **(1.7), ∇ is also decomposed into the form ∇ = ∇ H ∇V. **

(**[10]) The Chern connection on (M, Funiquely determined from the following conditions. **

(1) ∇ is *symmetric*:

**(1.15) **

where we considered π* = Σ(∂/∂*xi**dxi *.

(2) ∇ is *almost G-compatible*:

**(1.16) **

**Remark 1.8: **In the case of *C *= 0, the metric *F *is the norm function of a Riemannian metric *g*, and the Chern connection ∇ is given by ∇ = π* ∇*M *for the Levi-Civita connection ∇*M *of (*M, g*). The Chern connection is also called the *Rund connection *of (*M, F*) (cf. **[3], [12] [32]). **

We can easily show that θ defined by **(1.14) is invariant by the natural action m+. In local coordinate, θ is given by **

is the connection forms of ∇ with respect to {∂/∂*x*¹, …, ∂/∂*xn*}. The set {θ*i*, …, θ*n*is the dual basis of *V** defined by θ.

Then the covariant derivative ∇ is also decomposed into the form ∇ = ∇*H *∇*V*respectively. The covariant derivative ∇*G *of the metric *G *is decomposed into the form ∇*G *= ∇*H G *+ ∇*V G*, and thus the assumption **: **

By the definition **(1.4) of Cartan tensor field C, we have **

**(1.18) **

for all *X, Y, Z *∈ *A*).

On the other hand, **and **

Therefore we obtain

*Let **be the Ehresmann connection of π : TM× → M defined by ***(1.14) for the Chern connection ∇ of (M, F). Then we have **

**(1.19) **

Since the condition **. Then the condition (1.17) is written as dHG – tωG – Gare given by **

where (*Gij*) denotes the inverse of (*Gij*).

Let (*M, F*) be a Finsler manifold with the Chern connection ∇. For a non-vanishing vector field *v *= Σ*vi*(*x*)(∂/∂*xi*) on *M*, we define its covariant derivative ∇*v *be the natural lift of *v *. The covariant derivative ∇*v *:

If *v *satisfies ∇*v *= 0, then *v *is said to be *parallel *with respect to ∇.

Let *c *= (*x*(*t*)) : *I *= [0,1] → *M *be a smooth curve, and *v*(*t*) be a non-vanishing vector field along *c*is said to be *horizontal *:

If *v*(*t*) satisfies this equation, *v*(*t*) is said to be *parallel *along *c*.

The system **(1.22) has a unique solution vζ(t) which depends on the initial condition ζ = vζ(0) smoothly. From smooth dependence of solutions on ζ, the mapping Pc(t) : Tc(0)M → Tc(t)M defined by Pc(t)(ζ) = (c(t), vζ(t)) is a diffeomorphism for every t ∈ I. Because of homogeneity of θ, if vζ(t) is a solution of (1.22), then λvζ(t) is also a solution satisfying λvζ (0) = λζ, and thus the uniqueness of solutions implies that vλζ (t) = λvζ(tof a curve c starting at ζ ∈ Tc(0)M , and Pc(t) also satisfies the homogeneity Pc(t)(λζ) = λPc(t)(ζ) for all λ>0 and ζ ∈ Tc(0)M. The family Pc = {Pc(t) : t ∈ I} is called the parallel translation along c with respect to ∇. **

The tangent space *TxM *at every point *x *∈ *M **x *= *F*(*x*, •). If we put *Pc*(*t*)(ζ) = (*c*(*t*), *v*ζ(*t*)) for any point ζ in *Tc*(0)*M**v*ζ(*t*of the vector field *v*ζ(*t*) along *c*(*t*) is given by *F*(*c*(*t*), *v*ζ(*t*)). Then, because of **Proposition 1.9 we have **

Hence the parallel translation *Pc **Pc*(*t**c*(*t**c*(0).

*The parallel translation Pc along any curve c = c(t) on M is a norm-preserving map between the tangential normed-spaces*.

The parallel translation *Pc *is said to be *isometry *if it satisfies

**(1.23) **

for all ζ, η ∈ *TpM*. The parallel translation *Pc *along a curve *c *= *c*(*t*) is norm-preserving, but not isometry in general. It is trivial that, if *Pc *is a linear mapping, then *Pc *is an isometry. In a later section, we shall consider the case where every tangent spaces are isometric mutually as normed linear spaces.

We denote by *Cp *the set of all (piecewise) smooth curves *c *with starting point *p *= *c*(0) and ending point *p *= *c*(1). Then there exists a natural product ‘o’ in *Cp*. We also set *Hp *= {*Pc*(1) : *TpM *→ *TpM *| *c *∈ *Cp*}. Defining *Pc**Pc*2(1) = *P*(*c*1 · *c*2)(1), we can easily show that *Hp *is a group with the multiplication ‘o’. This group *Hp *is called the *holonomy group *with reference point *p *∈ *M*. In general, since the parallel translation *Pc *is not linear between the fibres, *Hp *is not a Lie group.

in the sequence **, and the Chern connection ∇ makes π* parallel by the assumption (1.15). We shall calculate the covariant exterior derivative ∇π*: **

Since *V *is integrable and π* satisfies **(1.12), we have ∇π*( XV, YV, we have **

and

Therefore we have

*The assumption ***(1.15) of the Chern connection ∇ is equivalent to **

**(1.24) **

*and *

**(1.25) **

*for all *.

by **(1.10). In Finsler geometry, the covariant exterior differential ∇θ plays an important role. **

The *torsion T *of ∇ is defined by

**(1.26) **

We remark that the vertical part of *T *vanishes identically. In fact, we have

, since the vertical subbundle *V *is integrable. Similar computations lead us to

*The horizontal part THH and the mixed part THV are given by *

*and *

**(1.28) **

*for all **respectively. In particular, the mixed part THV satisfies *

**(1.29) **

*where THH *(*X, Y*) = *T*(*XH*, *YH*) *and THV*(*X, Y*) = *T*(*XH*, *YV*).

**Remark 1.14: **Using local coordinate (*x*¹, · · ·, *xn*), the torsion form *T *is given by

**(1.30) **

**Case of: ***THH *≡ 0

The horizontal subbundle *H *is integrable if and only if [*H, H*] ⊂ *H*. Then, from **(1.13), H is integrable if and only if θ([XH, YH]) = 0 for all X, Y ∈ A). **

The *integrability tensor **of θ is defined by *

From **(1.27), we have THH = Θ, and thus THH = 0 means that the horizontal sub-bundle H is integrable. Therefore H on the total space TM× which is transversal to the fibres if THH ≡ 0. **

A non-vanishing smooth section *v *: *M *→ *TM*× is said to be *horizontal *with respect to ∇ if it satisfies ∇*v *= *v**∇ε = *v**θ = 0.

For a horizontal section *v*, we have *v***dV *≡ 0, and thus *v**dH *= *d **v**.

From this, the integrability condition *d*(*v**θ) = 0 for *v *to be horizontal is given by *v**θ = 0. Hence, if we assume Θ ≡ 0, this assumption guarantees the existence of a horizontal section *v*(*x*) = (*x, y*(*x*)) satisfying *v*(*x*0) = ζ(≠ 0) for an arbitrary initial point ζ ∈ *Tx*0*M*. The *n*-dimensional submanifold of *TM *defined by *y *= *y*(*x*through the point (*x*0, ζ).

For the metric *G *, we define a Riemannian metric *gv *on *M *by *gv *= *v***G *for a horizontal section *v*. For the connection form ω of ∇, the form ω*v *:= *v**ω defines a connection ∇*M *on *TM*. Then we have

*Assume that THH *≡ 0. *Let v : M *→ *TM*× *be a horizontal section, and gv the induced metric on M. The induced connection *∇*M on M is the Levi-Civita connection of (M, g), and v is parallel with respect to *∇*M*.

**Case of: ***THV *≡ 0 : **Landsberg spaces **

The metric *G *makes each fibre *TxM *a Riemannian space with the metric *Gx*. Since the parallel translation *Pc *along any curve *c *and its tangent vector field lies in horizontal space:

*Pc *is an isometry between the fibres if and only if

**(1.32) **

(cf. **[23]). We have then **

*The Lie derivative **XHG is given by *

*for all *.

A Finsler manifold (*M, F*) is said to be a *Landsberg *if *THV *≡ 0.

From the relation **(1.33), we can prove that (1.32) is equivalent to THV = 0. **

(**[22]) A Finsler manifold (M, F) is Landsberg if and only if the parallel translation Pc along any curve c is an isometry between the tangential Riemannian spaces. **

From this theorem, we see that any parallel translation *Pc *∈ *Hp *along *c *∈ *Cp *is a isometry in the tangential Riemannian space *TpM *if (*M, F*) is a Landsberg space. Then it is shown that *Hp *is a Lie group, since the isometry group *G *of any Riemannian manifold is a Lie group.

(**[26]) Suppose that (M, F) is a Landsberg space. Then the holonomy group Hp with reference point p ∈ M is a Lie group. **

Each fibre *TxM *is a Riemannian manifold with the metric induced form *G*. Then, the condition *THV *= 0 means that each fibre *TxM *is totally geodesic in *TM *with some Sasakian-type metric (cf. **[1]). **

On the other hand, the volume form on each fibre *TxM *is induced from the *n*. We shall consider the case where each fibre *TxM *is minimal in *TM*. From **(1.25), we have **

. If we put (tr *THV*)(*X*) := trace{*Y *→ *THV*(*X, Y*)}, then, by direct computations, we can show the following

*XH *Π = 0 if and only if (tr *THV*)(*X*) = 0.

, we denote by φ*t *the one parameter group of local transformation in *TM *induced from *XH*. For a compact subset *K*0 ⊂ *Tx*0*M*, *x*0 ∈ *M*, we set *Kt *= ϕ*t*(*K*0) ⊂ *T*ϕ*t*(*x*0)*M*. Then the volume *V*(*Kt*. Suppose that (*M, F*) is Landsberg space. Then by **Theorem 1.20, each ϕ t is isometry between the fibres, and we have **

Hence, if (*M, F*) is Landsberg, the volume *V*(*Kt*) is constant(cf. **[7]). **

A Finsler manifold (*M, F*) satisfying tr *THV *= 0 is called a *weak Landsberg space *(cf. **[44]). The condition tr THV = 0 means that each fibre TxM is is minimal submanifold in TM with some Sasakian-type metric (cf. [1]). **

An important quantity in geometry is the *curvature *which measures the flatness of the space.

The *curvature R *of ∇ is defined by

**(1.34) **

Similarly to the case of torsion *T*, we first remark that the vertical part of *R *vanishes identically. In fact, we obtain

, since the vertical subbundle *V *satisfies [*V, V*] ⊂ *V*. Hence the surviving part of *R *are the horizontal part *RHH *and the mixed part *RHV*:

and

The following is trivial from the definition.

**(1.35) **

From the assumption **(1.15), the Ricci identity ∇² π* = R ∧ π* gives R ∧ π* = 0. **

Then

and

induce the following.

*For all *, *the curvatures RHH and RHV satisfy the following identities: *

*and *

**(1.37) **

By the definition **(1.26), the Ricci identity ∇²ε = Rε implies the relation T = Rε. **

*The curvature R and the torsion T of the Chern connection *∇ *satisfy the relations *

**(1.38) **

*and *

**(1.39) **

*for all *.

The symmetry assumption **(1.15) derives Proposition 1.24. The almost G-compatibility assumption (1.16) derives the followings. Firstly, concerning with RHH, we have **

*The horizontal curvature RHH satisfies *

*for all X, Y, Z *.

Secondary, concerning the mixed part *RHV*, we have

*The mixed part RHV satisfies the following identities. *

*for all sections X, Y, Z and W of *.

As an application of **Proposition 1.27, we obtain the following **

*The mixed part RHV of the curvature R satisfies the following identity *

**(1.42) **

*for every X *∈ *A*).

**Remark 1.29: **The *curvature form *of ∇ is the 2-form on *TM*. By definition of *R*, we obtain the 1-st Bianchi identity Ω = dω + ω ∧ ω:

**(1.43) **

Differentiating this identity, we obtain the 2-nd Bianchi identity ∇Ω = 0:

In the case of Finsler geometry, this identity induces some complicated identities, since ∇ has non-zero torsions *THH *and *THV*. For example, if we calculate the horizontal 3-form of the left-hand-side of **(1.44), we obtain **

The other identities including the terms ∇*H RHV*, ∇*V RHH *and ∇*V RHV*, see the book **[10] or [32]. **

**Case of: ***RHH *≡ 0

We suppose *RHH *= 0. Then **(1.38) gives θ = THH = 0, and thus TM× admits a horizontal section v : M → TM×. **

*If RHH *= 0, *the induced metric gv *= *v***G is a flat Riemannian metric on TM, and so M is locally Euclidean*.

**Case of: ***RHV *= 0 : **Berwald spaces **

In this subsection, we shall consider the case of *RHV *= 0.

A Finsler manifold (*M, F*) is said to be *Berwald *if *RHV *= 0.

Because of **(1.39), a Berwald space is a special class of Landsberg spaces. If ( M, F) is Berwald, the Chern connection ∇ is linear, that is, there exists a symmetric linear connection ∇ on TM such that ∇ = π*∇M. **

If (*M, F*) is a Berwald space, then Szabó’s theorem **[49] showed that we can find a Riemannian metric g on M which is compatible with ∇M. We show the outline of the proof of this fact. For this, we define a isometric group G of F. For an arbitrary point x ∈ M, we set G = {g ∈ GL(ngy y , ∀y ∈ TxMand the homogeneity of F, we can prove that G is a compact Lie group [21]. Then we define an inner product (·, ·)x on TxM by **

for an arbitrary inner product (·, ·) on *TxM *and a bi-invariant Haar measure *dg *on G. By definition, it is trivial that this inner product 〈·, ·〉*x *is (G-invariant.

On the other hand, the holonomy group *Hx *of ∇*M *with the reference point *x *is a subgroup of G, since ∇*M *preserves *F *invariant. Hence 〈·, ·〉*x *is also *Hx*-invariant. Thus we can extend the inner product 〈·, ·〉*x *to a Riemannian metric *g *on *TM *by the help of the parallel displacement with respect to ∇*M*. It is trivial that ∇*M *is compatible with respect to this metric *g*. Hence we have

(**[49]) Suppose that a Finsler manifold (M, F) is a Berwald space. Then there exists a Riemannian metric g on M such that the Chern connection ∇ of(M, F) is given by ∇ = π*∇M for the Levi-Civita connection ∇M of(M,g). **

The curvature is related with parallel translation. The following theorem due to **[21] characterizes Berwald spaces in terms of parallel translations. **

(**[21]) A Finsler manifold (M, F) is Berwald if and only if the parallel translation Pc along any curve c is an isometry between the tangential normed spaces. **

**Proof: **We suppose that the parallel translation *Pc *along any curve *c *= *c*(*t*) in *M *is an isometry between the tangential normed spaces. Then, by the Mazur-Ulam’s theorem (cf. **[33] and [50]), the mapping Pc is linear and there exists a GL(nsatisfying **

. From **(1.22) we get **

for all curves *c*(*t*are linear with respect to the fibre coordinate (*y*¹, · · ·, *yn*, which shows *RHV *= 0.

Conversely we suppose that (*M, F*are linear in the fibre coordinate (*y*¹, · · ·, *yn*), the solutions *y*ζ(*t*) of **(1.22) are linear in ζ. Hence the parallel translation Pc is linear, and from (1.23), we have **

for all ζ, η ∈ *Tc*(0) *M*. Hence *Pc *is an isometry.

Let *F *= α + β be a Randers metric on *M*. Then, by the well-known theorem due to **[25], ( M, F) is Berwald if and only if β is parallel 1-form on the base Riemannian manifold (M, g). **

**Remark 1.35: **From **(1.39), if RHV vanishes identically, then THV also vanishes. Hence the class of Landsberg spaces contains the class of Berwald spaces. There exists a lot of example of Berwald spaces. However it is still an open problem to find an example of non-Berwald Landsberg space. **

**Case of: ***RHH *= *RHV *= 0 : **Locally Minkowski spaces **

In this section, we shall be concerned with flat Finsler manifolds.

A Finsler manifold (*M, F*) is said to be *locally Minkowski *if there exists a local coordinate system on *M *with respect to which the function *F *is independent of the base point *x *∈ *M*.

We have the following well-known theorem (cf. **[32]). **

*A Finsler manifold (M, F) is locally Minkowski if and only if the Chern connection *∇ *is flat*.

**Proof: **We suppose that ∇ is flat, that is, *RHH *= *RHV *= 0. Then, the Chern connection ∇ is induced from a flat connection ∇*M *on *TM*. Hence there exists an open cover U of *M *and local frame fields (*e*1, · · ·, *en*) on *U *∈ U such that ∇*ej *= 0. This condition is equivalent to the existence of the change of frames

on each *U *satisfying *dA *+ ω*A *= 0. With respect to such a local frame field *e*of ∇ vanishes on *U*and **(1.19) imply the independence of F on the base point x ∈ Uthe inverse of the matrix A, the condition above is equivalent to **

Since ∇ is symmetric, there exist some functions *wi*(*x*, and the local frame *ej *is given by

Hence there exists a local coordinate system {*U*, (*wi*)} on *M *with respect to which the function *F *is independent on the base point *x *∈ *U*.

Conversely we assume that *F *is independent of the base point *x*. By definition, the metric tensor *Gij *is also independent of the base point *x*and **and thus the Chern connection ∇ is flat. **

Let *F *= α + β be a Randers metric on *M*. Then (*M, F*) is locally Minkowski if and only if (*M, F*) is Berwald and the base Riemannian manifold (*M, g*) is flat (**[25]). **

This example is true for any locally Minkowski space.

*A Finsler manifold (M, F) is locally Minkowski if and only if (M, F) is Berwald and its associated Riemannian metric is flat*.

Let *X *be a tangent vector at *x *∈ *M*. We may consider *X *. Then the 2-plane *F*(*X*) spanned by *X *and ε is called the *flag *with the *flagpole *ε. For the curvature tensor *R *of ∇, the sectional curvature

is called *the flag curvature *of the flag *F*(*X*), and denoted by *K*. From **(1.42), we have G(RHV(X, ε) ε, X, and so the flag curvature K is given by **

**(1.45) **

The flag curvature *K *depends on *X *and the point (*x, y*) ∈ *TM*× : *K *= *K*(*x, y, X*). If the flag curvature *K *is independent of *X *at every point (*x, y*) ∈ *TM*×, the space is said to be *of scalar flag curvature K*(*x, y*). A Finsler manifold (*M, F*) is said to be of *constant flag curvature *if *K *is constant. For Finsler manifolds of constant flag curvature, see Chapter 12 in **[10]. The proof of the following theorem is found in [32]. **

(Schur’s lemma) *Let (M, F) be a Finsler manifold of scalar flag curvature K = K(x, y). If K is a function of position x *∈ *M alone, then (M, F) is of constant flag curvature provided *dim *M *≥ 3.

Let *F *be the Funk metric on the unit ball B defined in Example 1.1. It is well-known that *F *has negative constant flag curvature *K *= −1 (see **[37] or [14]). **

Let γ : *I *= [0,1] → *M *be a smooth curve. Since the symmetry condition *F*(*x, y*) = *F*(*x, −y*) is not assumed, the orientation of curves is essential, that is, if a curve γ is given, then we always assume that γ is *oriented *by the parameter *t*. A smooth curve γ = γ(*t*) is said to be *regular *for every *t *∈ *I*.

Let Γ(*p, q*) be the set of all regular oriented curves with the initial point *p *= γ(0) and the terminal point *q *= γ(1). Then we define a functional *LF *: Γ(*p, q*by

Since *F *satisfies the homogeneity condition, this definition is well-defined. For an ordered pair (*p, q*) ∈ *M *× *M*, the distance function *dF*(*p, q*) is defined by *dF*(*p, q*) = inf*c LF*(γ), where infinimum is taken over of all oriented (piecewise) smooth curves from *p *to *q*. In general, since the symmetry condition is not assumed, the distance function *df *does not satisfy the symmetric property *dF*(*p, q*) = *dF*(*q, p*). However, the distance *dF *satisfies the following conditions:

(1) *dF*(*p, q*) ≥ 0,

(2) *dF*(*p, q*) = 0 if and only if *p = q*,

(3) *dF*(*p, q*) ≤ *dF*(*p, r*) + *dF*(*r, q*).

The metric topology of *M *is defined by the sets *B*(*p*, δ) = {*q *∈ *M *| *dF*(*p, q*)<δ}, and the metric topology of a connected Finsler manifold (*M, F*) coincides with the manifold topology of *M*.

The *canonical lift *For a vector field *X*(*t*) along γ, we consider *X*(*t*) as a section of *TM *. Then we use the notation ∇*tX *:

, since *X *is a vector field along the curve γ(*t*) in the base manifold *M*. The definition **(2.1) and the metrical condition (1.14) imply **

for all vector fields *X*(*t*) and *Y*(*t*) along γ. Hence we have

If ∇*tX *= ∇*tY *= 0, *then the inner product G(X, Y) is constant along *γ.

Let (*M, F*) be a Finsler manifold with the Chern connection ∇.

A regular oriented curve γ : *I *→ *M *is said to be a *path *.

The length *s *, and the function *s*(*t*) is an increasing function of the parameter *t*. If the parameter *t *is positively proportional to *s*, then *t *is said to be *normal*.

Let (*M, F*) be a Finsler manifold. A path in *M *with a normal parameter is called a is a *geodesic *in (*M, F*).

From **(2.1), a regular oriented curve γ( t) = (xi(t)) with normal parameter t : **

**(2.3) **

is satisfied. From the metrical condition **(2.2), if γ = γ( t) with normal parameter t has a constant norm and γ has constant speed. In the sequel, we always assume that the parameter of a geodesic to be normal otherwise stated. **

Let γ*X *: *I *→ *M *be a geodesic with initial point *x *= γ*X*, where the parameter *t *is, of course, is normal. We shall define the *exponential map *exp by exp (*x, X*) = γ*X*(1) if *X *= 0 and exp (*x*, 0) = *x*. The restriction of exp to D ∩ *TxM *is denoted by exp*x*. The restricted exponential map exp*p *maps the rays through the origin 0 ∈ *TxM *to the unique geodesics through the point *x *in sufficiently small *Bx*(*r*) = {*X *∈ *TxM **X *<*r*}.

The exponential map exp is defined on an open neighborhood D of the zero section *z*(*M*) of *TM*, and exp is *C*∞-class away from *z*(*M*). Furthermore exp is *C*¹-class at *z*(*M*), and its derivative at *z*(*M*) is the identity map. By a result due to Akbar-Zadeh, the map exp is *C*²-class at *z*(*M*) if and only if (*M, F*) is a Berwald space (see **[10]). **

For each *X *∈ *TxM*, the *radial geodesic *γ*X *is given by γ*X*(*t*) = exp*x*(*tX*) for all *t *∈ *I *such that either side is defined. This geodesic segment γ*X *, **is constant along γ X**

In this section, we shall show the first variation formula in Finsler manifolds. For this end, we introduce some definitions.

Let γ = γ(*t*) ∈ Γ(*p, q*. Then a *variation *of γ is a family {γ*s*} of oriented curves γ*s*(*t*) parameterized by *s *∈ (-ε, ε) such that γ0(*t*) = γ(*t*) for all *t *∈ *I*. A variation Γγ is said to be *proper *if it fixes the end points, that is, γ*s*(0) = *p *and γ*s*(1) = *q*. We suppose that the map Γγ : (-ε, ε) × *I *→ *M *defined by Γγ(*s,t*) = γ*s*(*t*) is smooth. (For the variational problem of arc length, it is enough to assume that Γγ is piecewise differentiable with respect to the parameter *t *(cf. **[32], Chapter VII), however, we shall assume the smoothness of Γγ for the simplicity of discussions.) **

By the assumption, the map Γ satisfies Γγ(0, *t*) = γ(*t*), *p *= Γγ(*s*, 0) and *q *= Γγ(*s*, 1). Setting *s *= constant for each *s *∈ (-ε, ε), the parameterized curve γ*s *: *I *→ *M *defined by γ*s*(*t*) = Γγ(*s, t*) is called a *s-curve*, while the parameterized curve γ*t*(*s*) = Γγ(*s,t*) is a *t-curve *which is transversal curve to γ. In local coordinates, we set Γγ(*s,t*) = (*x*¹(*s,t*), · · ·, *xn*(*s, t*)). We denote by S = ∂γ*t*/∂*s *and T = ∂γ*s*/∂*t *the tangent vector fields of *t*-and *s*-curves respectively:

In particular, the vector field V(*t*) along γ defined by

is called the *variational field *induced from Γγ. If Γγ is proper, that is, Γγ satisfies γ*s *(0) = γ(0) = *p *and γ*s *(1) = γ(1) = *q *for all *s *∈ (-ε, ε), then the variational field V is *proper*, that is, V satisfies V(0) = V(1) = 0.

at least one point on γ. Let V = V(*t*) be any vector filed along a regular oriented curve γ = γ(*t*). Then there exists a variation Γγ which induces V as its variational field. In fact, if we take Γγ(*s, t*) = exp(*s*V(*t*)), then Γγ : (-ε, ε) × *I *→ *M *is a variation of γ with variational field V.

*Let *V *be any vector field along *γ. *Then *V *is a variational field of some variation *Γγ *of *γ. *If *V *is proper, then *V *is the variational field induced from a some proper variation *Γγ.

of *s*-curve γ*s*:

*Let *Γγ : (-ε, ε) × *I *→ *M be a variation. Then we have *

**(2.4) **

*along *

Let Γγ be a proper variation of a regular oriented curve γ ∈ Γ(*p, q*). We compute the first variation of length functional *LF*(γ*s*). Since

, we have

Furthermore, **(1.17) and (2.4) imply **

. Consequently we have

which gives

Evaluating *s *derives the following:

(First Variation Formula) *Let *γ : *I *→ *M be a regular oriented curve, and *Γγ *a proper variation of *γ. *Then *

**(2.6) **

*where *V *is the variational field of *Γγ.

A regular oriented curve γ is said to be a *stationary point *of the functional *LF *if (*dLF*(γ*s*)/*ds*)*s*=0 = 0 for any proper variation Γγ. If a regular oriented curve γ : *I *→ *M *is a geodesic, then γ satisfies **(2.3), and thus γ is a stationary point of LF from (2.6). **

Conversely we suppose that γ is a stationary point of the functional *LF*. Since the condition (*dLF*(γ*s*)/*ds*)*s*for a smooth function ϕ satisfying ϕ(0) = ϕ(1) = 0 and ϕ>0 elsewhere. Then, since V is proper and from **(2.6), we have **

on *I*.

*A regular oriented curve in a Finsler manifold (M, F) is a stationary point of the functional LF if and only if *γ *is a geodesic from p to q*.

A curve γ from *p *= γ(0) to *q *= γ(1) is said to be *minimizing *if *dF*(p, q) = *LF*(γ). Since minimizing curve is a stational curve, we have

*Every minimizing curve in (M, F) is a geodesic if *γ *is regular*.

The converse of this theorem is also true.

*Every geodesic in a Finsler manifold (M, F) is locally minimizing*.

This theorem is proved by using the Gauss lemma. We define the *geodesic ball **x*(*r*) centered at *x *∈ *M *of radius *r **x*(*r*) = exp (*Bx*(*r*)) for the tangential ball *Bx*(*r*) = {ζ ∈ *TxM *<*r*}. Let *Sx*(*r*) = {*X *∈ *TxM **X *= *r*} be the tangent sphere. Then the set *Sx*(*r*) = exp (*Sx*(*r*)) is called the *geodesic sphere *at *x *of radius *r*. Then the Gauss lemma is stated as follows.

(The Gauss Lemma) *The radial geodesic *γ*X is orthogonal to the geodesic sphere Sx*(*r*) at *x *∈ *M*.

For the proof of **Theorem 2.9, we need more technical preliminaries, and thus we omit it here. For the complete proof, see [10] or [13]. **

From Proposition 2.8, a geodesic in (*M, F*) is characterized as the critical points of length functional *LF*. In general, the equation which characterizes the critical points of a functional on Γ(*p, q*) is called the *Euler-Lagrange equation *of *LF*. For details, refer to the book **[4] or [32]. **

We consider an arbitrary proper variation Γγ : γ*s*(*t*) = γ(*t*) + *sX*(*t*) of a smooth curve γ(*t*) with fixed end points, that is, *X*(0) = *X*. The Taylor extension gives

and this extension implies

Because of

we have

Hence, γ(*t*) is a critical point of *LF *if and only if

**(2.7) **

This equation is the *Euler-Lagrange equation *of the functional *LF*. The quantity *Ei *is written as

where we put

Putting *Gi *= Σ *Gim Gm *and *yi *= *dxi*/*dt*, the Euler-Lagrange equation *Ei*(γ) = 0 implies the equation of the geodesic as follows:

for the function *p *= *d*log*F/dt*, where we note that *t *is an arbitrary parameter of γ. In particular, if we take Finslerian arc length as the parameter *t *of γ, that is, *dt = F*(*x, dx*), we obtain the equation of geodesic as follows:

**(2.8) **

(*Geodesics of Randers metrics*) Let *F *= α+β be a Randers metric, where α² = Σ *gij *(*x*)*yiyj *and β = Σ *bi*(*x*)*yi*the Chiristoffel symbol of the base Riemannian manifold (*M, g*). Let *t *be the Finslerian arc length.

by

Then, by direct computations, we see that the functions *Gi *in **(2.8) are given by **

Hence the equation **(2.8) with the Finslerian arc length parameter t is given by the following complicated form: **

If we take the Riemannian arc length *u *as the parameter, that is, *du *= α(*x, dx*), we have *dt = du*+Σ *bi*(*x*)*dxi*, and so

Then the equation above is reduced to the following form:

In particular, if the 1-form β = Σ *bi*(*x*)*dxi *implies

and so any geodesic in (*M, F*) coincides with one in the base Riemannian manifold (*M, g*). Consequently it is shown that, if the 1-form β is closed, then (*M, F*) is *projectively equivalent *to the base Riemannian manifold (*M, g*).

(*Geodesics of Funk metric*) Let *F **n *stated in Example 1.1. In this case, the 1-form β is exact form, and so any geodesic in (B, *F*) is given by the one in the Hilbert’s space (B, *gH*). We shall show the equation of geodesic parameterized by its Finslerian arc length *t*. From Example 2.1, it is given by

Since the Hilbert’s metric *gH *= Σ *gij *(*x*)*dxi **dxj *is given by

is given by

Then, because of

we obtain

Hence the equation **(2.8) is given by **

Furthermore, from

, we obtain

Consequently, the equation of geodesics in (B, *F*) is given by the following simple form:

The solution of this differential equation with initial conditions *A *= (*a*¹, · · ·, *an*) = *x*is given by *xi *= λ*i*(1 – *e*−*t*)+*ai*. Therefore any geodesic in (B, *F*) is given by a line in B.

Let *P *the point (λ¹+*a*¹, · · ·, λ*n*+*an*) = *x*(∞) on the boundary ∂B, and *B *a point on the line. We denote by *AP *(resp. *BP*) the Euclidean distance between the points *A *and *P *(resp. *B *and *P*). Because of

which implies that the Funk’s distance *df *is given by

Since the Hilbert’s distance *dH *is given by *dH*(*A, B*) = [*dF*(*A, B*)+*dF*(*B, A*)]/2, the distance *dH *is given by

A variation Γγ = Γγ(*s,t*) of a geodesic γ is said to be a *geodesic variation *if each *s*-curve γ*s *is also a geodesic. Since each *s*-curve γ*s *.

Let *X *be a vector field along γ*s*. Then, since [*S*, T] = 0, we have

along γ*s*. From this equation, we get the so-called *the Jacobi equation*.

*(The Jacobi Equation) Let *γ *be a geodesic, and *V *the variational field of a geodesic variation *Γγ *of *γ *in a Finsler manifold (M, F). Then *V *satisfies *

**(2.10) **

Let (*M, F*) be a Finsler manifold. The differential equation **(2.10) is called the Jacobi equation. A vector field J along a geodesic satisfying (2.10): **

is called a *Jacobi field *in (*M, F*).

By definition, the variational field V of a geodesic variation of a geodesic γ is a Jacobi field. Conversely, every Jacobi field along a geodesic γ is the variational field of some geodesic variation of γ. The differential equation **(2.10) is linear and of second order, we have 2 n linearly independent solution. Therefore, along any geodesic γ, the set of Jacobi field is a 2n-dimensional vector space. **

Let γ ∈ Γ(*p, q*) be a geodesic segment in *M*. Then *q *is said to be *conjugate along *γ if there exists a Jacobi field *J*(≠ 0) along γ such that *J *vanishes at *p *and *q*.

For *X *∈ *TpM*, we set *q *= exp*p X*. For an arbitrary *Y *∈ *TX *(*TpM*), we shall compute the differential (exp*p*)**Y *at *X*:

To compute (exp*p*)*, we define a geodesic variation Γγ of γ*X *by Γγ(*s, t*) = exp*p t*(*X*+*sY*). The variational field *J *= ∂ Γγ/∂*s *is a Jacobi field along γ*X*, and we have *J*(1) = (exp*p*)**Y*. The conjugate points are the image of the singularities by the exponential mapping.

*Let *γ*X*(*t*) = exp*p*(*tX*) (*t *ε *I*) *be the radial geodesic for X *ε *TxM*. *Then *exp*p is a local diffeomorphism if and only if q *= exp*p X is not conjugate to p along *γ*X*.

Let γ : *I *→ *M *be a geodesic with unit speed. We shall compute the second variation of the length functional *LF*. We shall compute

Differentiating with respect to *s*, we have

From **(2.2) and (2.4), we get **

Furthermore

Consequently we have

along γ*s*and V(0) = V(1) = 0 imply

we have

and ∇*t *V⊥ = ∇*t *V – ∇*t *= (∇*t *V)⊥. Hence we have

from **from (1.36). Hence we get **

Consequently, we obtain the *second variation formula of LF*.

(Second Variation Formula) *Let *γ : *I *→ *M be any geodesic with unit speed, and *Γγ *a proper variation of *γ, *and *V *its variation field. Then *

*where *V⊥ *is the normal part of *V.

Since **(1.36) implies **

, we obtain

for the flag curvature *K*. Hence the second variation formula **(2.12) has the form **

Therefore we have

*Let (M, F) be a Finsler manifold with non-positive flag curvature K. Then, the second variation of any geodesic satisfies *

We define the *index form *on a Finsler manifold (*M, F*). Let γ be a unit speed geodesic in (*M, F*). We set

for normal proper vector fields *X, Y *along γ. The index form *I *is a symmetric bi-linear form on the space of normal proper vector fields. In fact, the Bianchi identity **(1.36) implies **

Since the last term on the left hand side vanishes from **(1.6) and (1.42), we have **

along γ. Thus *I *is a symmetric bi-linear form: *I*(*X, Y*) = *I*(*Y, X*).

Since **(1.35) induces **

along γ, if *X *and *Y *are proper, we have

which implies

By the definition of *I *and **(2.11), the second variation of Lp of unit speed geodesic is given by I(X, X), and it can be thought as the Hessian of the length functional LF. Thus, if γ is minimizing, then I(X, X) ≥ 0 for any proper normal vector field X along γ. The following is a generalization of the well-known theorem in Riemannian geometry which shows that no geodesics is minimizing past its first conjugate point (e.g., Theorem 10.15 in [30]). **

*If *γ ε Γ(*p, *

Close Dialog## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

Loading